MINE: Mutual Information Neural Estimation

MINE: Mutual Information Neural Estimation. Belghazi, M. I., Baratin, A., Rajeswar, S., Ozair, S., Bengio, Y., Courville, A., & Hjelm, R. D.

Paper abstract bibtex

We argue that the estimation of mutual information between high dimensional continuous random variables can be achieved by gradient descent over neural networks. We present a Mutual Information Neural Estimator (MINE) that is linearly scalable in dimensionality as well as in sample size, trainable through back-prop, and strongly consistent. We present a handful of applications on which MINE can be used to minimize or maximize mutual information. We apply MINE to improve adversarially trained generative models. We also use MINE to implement Information Bottleneck, applying it to supervised classification; our results demonstrate substantial improvement in flexibility and performance in these settings.

@article{belghaziMINEMutualInformation2018,
  archivePrefix = {arXiv},
  eprinttype = {arxiv},
  eprint = {1801.04062},
  primaryClass = {cs, stat},
  title = {{{MINE}}: {{Mutual Information Neural Estimation}}},
  url = {http://arxiv.org/abs/1801.04062},
  shorttitle = {{{MINE}}},
  abstract = {We argue that the estimation of mutual information between high dimensional continuous random variables can be achieved by gradient descent over neural networks. We present a Mutual Information Neural Estimator (MINE) that is linearly scalable in dimensionality as well as in sample size, trainable through back-prop, and strongly consistent. We present a handful of applications on which MINE can be used to minimize or maximize mutual information. We apply MINE to improve adversarially trained generative models. We also use MINE to implement Information Bottleneck, applying it to supervised classification; our results demonstrate substantial improvement in flexibility and performance in these settings.},
  urldate = {2019-04-01},
  date = {2018-01-12},
  keywords = {Statistics - Machine Learning,Computer Science - Machine Learning},
  author = {Belghazi, Mohamed Ishmael and Baratin, Aristide and Rajeswar, Sai and Ozair, Sherjil and Bengio, Yoshua and Courville, Aaron and Hjelm, R. Devon},
  file = {/home/dimitri/Nextcloud/Zotero/storage/3FZITPZ2/Belghazi et al. - 2018 - MINE Mutual Information Neural Estimation.pdf;/home/dimitri/Nextcloud/Zotero/storage/VM34ZB9U/1801.html}
}

Downloads: 0

{"_id":"QmN8CPxxCfQtoK3PE","bibbaseid":"belghazi-baratin-rajeswar-ozair-bengio-courville-hjelm-minemutualinformationneuralestimation","authorIDs":[],"author_short":["Belghazi, M. I.","Baratin, A.","Rajeswar, S.","Ozair, S.","Bengio, Y.","Courville, A.","Hjelm, R. D."],"bibdata":{"bibtype":"article","type":"article","archiveprefix":"arXiv","eprinttype":"arxiv","eprint":"1801.04062","primaryclass":"cs, stat","title":"MINE: Mutual Information Neural Estimation","url":"http://arxiv.org/abs/1801.04062","shorttitle":"MINE","abstract":"We argue that the estimation of mutual information between high dimensional continuous random variables can be achieved by gradient descent over neural networks. We present a Mutual Information Neural Estimator (MINE) that is linearly scalable in dimensionality as well as in sample size, trainable through back-prop, and strongly consistent. We present a handful of applications on which MINE can be used to minimize or maximize mutual information. We apply MINE to improve adversarially trained generative models. We also use MINE to implement Information Bottleneck, applying it to supervised classification; our results demonstrate substantial improvement in flexibility and performance in these settings.","urldate":"2019-04-01","date":"2018-01-12","keywords":"Statistics - Machine Learning,Computer Science - Machine Learning","author":[{"propositions":[],"lastnames":["Belghazi"],"firstnames":["Mohamed","Ishmael"],"suffixes":[]},{"propositions":[],"lastnames":["Baratin"],"firstnames":["Aristide"],"suffixes":[]},{"propositions":[],"lastnames":["Rajeswar"],"firstnames":["Sai"],"suffixes":[]},{"propositions":[],"lastnames":["Ozair"],"firstnames":["Sherjil"],"suffixes":[]},{"propositions":[],"lastnames":["Bengio"],"firstnames":["Yoshua"],"suffixes":[]},{"propositions":[],"lastnames":["Courville"],"firstnames":["Aaron"],"suffixes":[]},{"propositions":[],"lastnames":["Hjelm"],"firstnames":["R.","Devon"],"suffixes":[]}],"file":"/home/dimitri/Nextcloud/Zotero/storage/3FZITPZ2/Belghazi et al. - 2018 - MINE Mutual Information Neural Estimation.pdf;/home/dimitri/Nextcloud/Zotero/storage/VM34ZB9U/1801.html","bibtex":"@article{belghaziMINEMutualInformation2018,\n archivePrefix = {arXiv},\n eprinttype = {arxiv},\n eprint = {1801.04062},\n primaryClass = {cs, stat},\n title = {{{MINE}}: {{Mutual Information Neural Estimation}}},\n url = {http://arxiv.org/abs/1801.04062},\n shorttitle = {{{MINE}}},\n abstract = {We argue that the estimation of mutual information between high dimensional continuous random variables can be achieved by gradient descent over neural networks. We present a Mutual Information Neural Estimator (MINE) that is linearly scalable in dimensionality as well as in sample size, trainable through back-prop, and strongly consistent. We present a handful of applications on which MINE can be used to minimize or maximize mutual information. We apply MINE to improve adversarially trained generative models. We also use MINE to implement Information Bottleneck, applying it to supervised classification; our results demonstrate substantial improvement in flexibility and performance in these settings.},\n urldate = {2019-04-01},\n date = {2018-01-12},\n keywords = {Statistics - Machine Learning,Computer Science - Machine Learning},\n author = {Belghazi, Mohamed Ishmael and Baratin, Aristide and Rajeswar, Sai and Ozair, Sherjil and Bengio, Yoshua and Courville, Aaron and Hjelm, R. Devon},\n file = {/home/dimitri/Nextcloud/Zotero/storage/3FZITPZ2/Belghazi et al. - 2018 - MINE Mutual Information Neural Estimation.pdf;/home/dimitri/Nextcloud/Zotero/storage/VM34ZB9U/1801.html}\n}\n\n","author_short":["Belghazi, M. I.","Baratin, A.","Rajeswar, S.","Ozair, S.","Bengio, Y.","Courville, A.","Hjelm, R. D."],"key":"belghaziMINEMutualInformation2018","id":"belghaziMINEMutualInformation2018","bibbaseid":"belghazi-baratin-rajeswar-ozair-bengio-courville-hjelm-minemutualinformationneuralestimation","role":"author","urls":{"Paper":"http://arxiv.org/abs/1801.04062"},"keyword":["Statistics - Machine Learning","Computer Science - Machine Learning"],"downloads":0},"bibtype":"article","biburl":"https://raw.githubusercontent.com/dlozeve/newblog/master/bib/all.bib","creationDate":"2020-01-08T20:39:39.312Z","downloads":0,"keywords":["statistics - machine learning","computer science - machine learning"],"search_terms":["mine","mutual","information","neural","estimation","belghazi","baratin","rajeswar","ozair","bengio","courville","hjelm"],"title":"MINE: Mutual Information Neural Estimation","year":null,"dataSources":["3XqdvqRE7zuX4cm8m"]}